{"id":38,"date":"2026-04-08T06:14:41","date_gmt":"2026-04-08T06:14:41","guid":{"rendered":"https:\/\/karloneill.dev\/?p=38"},"modified":"2026-04-08T06:28:35","modified_gmt":"2026-04-08T06:28:35","slug":"why-did-the-banana-go-to-the-doctor-getting-better-answers-out-of-generative-ai","status":"publish","type":"post","link":"https:\/\/karloneill.dev\/?p=38","title":{"rendered":"Why did the banana go to the doctor? Getting better answers out of generative AI"},"content":{"rendered":"\n

“Why did the banana go to the doctor?”
Because he wasn’t peeling well!<\/strong> <\/p>\n\n\n\n

That was the identical answer given when I asked “Tell me a joke about a banana” by ChatGPT, Claude and Gemini. My 4 year old grand-daughter told me that joke, so is AI intelligent? It is not. AI is as dumb as a post. Surely there were some variations between each of the generative AI’s.<\/p>\n\n\n\n

The purpose of my doing this experiment is not to give you an arsenal of banana jokes. It is to demonstrate that by asking questions in a different way you can get a lot more variation in answers from generative AI. It will lead you to getting a lot better use out of it. Use the same approach when you are doing your office work and looking for AI to give you some guidance, ideas or getting it to do something for you.<\/p>\n\n\n\n

CHATGPT<\/strong><\/p>\n\n\n\n

Now try this in ChatGPT:<\/p>\n\n\n\n

“Tell me 5 jokes about a banana with probabilities”<\/em><\/strong><\/p>\n\n\n\n

ChatGPT’s answer (you may get a different response hopefully):<\/p>\n\n\n\n

Here are 5 banana jokes, each with a totally made-up \u201cprobability of making you laugh\u201d:<\/p>\n\n\n\n

    \n
  1. Why did the banana go out with the prune?<\/strong>
    Because it couldn\u2019t find a date<\/em>!
    Probability of laughter:<\/strong> 72%<\/li>\n\n\n\n
  2. What do you call two banana peels on the floor?<\/strong>
    A slippery situation<\/em>.
    Probability of laughter:<\/strong> 64%<\/li>\n\n\n\n
  3. Why don\u2019t bananas ever feel lonely?<\/strong>
    Because they hang out in bunches<\/em>.
    Probability of laughter:<\/strong> 78%<\/li>\n\n\n\n
  4. Why did the banana fail the test?<\/strong>
    It split<\/em> under pressure.
    Probability of laughter:<\/strong> 69%<\/li>\n\n\n\n
  5. What\u2019s a banana\u2019s favorite gym exercise?<\/strong>
    The split<\/em>.
    Probability of laughter:<\/strong> 75%<\/li>\n<\/ol>\n\n\n\n

    If you want, I can crank those probabilities up to 99%\u2026 but that might require better jokes \ud83d\ude04<\/p>\n\n\n\n

    \n\n\n\n

    See how framing the question differently to ChatGPT sets it free from a standard answer and gives you something to work with. I genuinely think my grand-daughter will find some of those funny. It even showed a bit of what might be called “personality”.<\/p>\n\n\n\n

    \n\n\n\n

    GEMINI<\/strong><\/p>\n\n\n\n

    Now lets try with Gemini:<\/p>\n\n\n\n

    My prompt:<\/p>\n\n\n\n

    “Tell me 5 jokes about a banana with probabilities”<\/em><\/strong><\/p>\n\n\n\n

    Gemini’s answer:<\/p>\n\n\n\n

    Since we\u2019re mixing fruit with the laws of chance, let’s look at the statistical likelihood of these landing. Here are five banana jokes, ranked by their mathematical probability of making you exhale slightly faster through your nose.<\/p>\n\n\n\n

    \n\n\n\n

    1. The Potassium Paradox<\/h3>\n\n\n\n

    The Joke:<\/strong> Why did the banana go to the doctor? Because it wasn’t peeling<\/strong> very well.<\/p>\n\n\n\n

      \n
    • Probability of a groan:<\/strong> P(Groan) = 0.98<\/li>\n\n\n\n
    • Likelihood of being told at a BBQ:<\/strong> High (85% confidence interval).<\/li>\n<\/ul>\n\n\n\n

      2. The Slip-up Distribution<\/h3>\n\n\n\n

      The Joke:<\/strong> What is the probability of a banana skin causing a fall?<\/p>\n\n\n\n

      The Answer:<\/strong> It depends on the coefficient of friction<\/strong>, but the results are usually slippery<\/strong> at best.<\/p>\n\n\n\n

        \n
      • Probability of physical comedy:<\/strong> P(Slip) approx 1 – P(Balance)<\/li>\n\n\n\n
      • Statistical Significance:<\/strong> p < 0.05 (Significant bruising).<\/li>\n<\/ul>\n\n\n\n

        3. The Time-Series Analysis<\/h3>\n\n\n\n

        The Joke:<\/strong> Why do bananas use sunscreen? Because they peel<\/strong> if they spend too much time in the sun.<\/p>\n\n\n\n

          \n
        • Probability of sunburn:<\/strong> $P(Burn | UV Exposure > Threshold) = 1.0<\/li>\n\n\n\n
        • Reliability:<\/strong> This joke has a high Standard Deviation<\/strong> from actually being funny.<\/li>\n<\/ul>\n\n\n\n

          4. The Bayesian Banana<\/h3>\n\n\n\n

          The Joke:<\/strong> A banana is lying on the ground. A statistician walks by and says, “Given that I see a banana, there is a 99% probability that someone is about to have a comedy trope happen to them.”<\/p>\n\n\n\n

            \n
          • Prior Probability:<\/strong> $P(Prank) = 0.50<\/li>\n\n\n\n
          • Posterior Probability:<\/strong> $P(Prank | Banana on Floor) = 0.99<\/li>\n<\/ul>\n\n\n\n

            5. The Discrete Variable<\/h3>\n\n\n\n

            The Joke:<\/strong> Why did the banana fail the math test? Because it couldn’t find the squared<\/strong> root\u2014it only understood curved<\/strong> geometry.<\/p>\n\n\n\n

              \n
            • Probability of accuracy:<\/strong> 100% (Bananas are non-Euclidean).<\/li>\n\n\n\n
            • Confidence Level:<\/strong> 95% that the teacher didn’t find it “appealing.”<\/li>\n<\/ul>\n\n\n\n
              \n\n\n\n
              \n

              Note:<\/strong> The margin of error for these jokes is plus or minus 15%, depending on how much you actually like puns.<\/p>\n<\/blockquote>\n\n\n\n

              \n\n\n\n

              OK. This time Gemini misunderstood me altogether and made jokes about bananas and<\/em> probability; only understandable by mathematicians. I have to reframe the question in a new chat. The reason for putting it in a new chat is to avoid the previous answer being taken into account. AI is dumb, it won’t remember anything in a new chat.<\/p>\n\n\n\n

              “Tell me 5 jokes about a banana and include the probabilities of whether people are likely to find the jokes funny.”<\/em><\/strong><\/p>\n\n\n\n

              Gemini’s answer this time:<\/p>\n\n\n\n

              Here are 5 banana jokes, complete with highly scientific (okay, entirely made-up but realistic) probabilities of how likely people are to find them funny:<\/p>\n\n\n\n

              The Banana Comedy Club<\/h3>\n\n\n\n

              1. The Classic Dad Joke<\/strong><\/p>\n\n\n\n

                \n
              • Joke:<\/strong> Why did the banana go to the doctor? Because he wasn’t peeling well!<\/li>\n\n\n\n
              • Probability of finding it funny:<\/strong> 85%<\/strong><\/li>\n\n\n\n
              • The Breakdown:<\/strong> This is the gold standard of fruit comedy. It\u2019s a reliable, groan-inducing dad joke that hits the pun perfectly. You’ll get plenty of eye rolls, but deep down, they’re laughing.<\/li>\n<\/ul>\n\n\n\n

                2. The Spooky Pun<\/strong><\/p>\n\n\n\n

                  \n
                • Joke:<\/strong> What is a ghost’s favorite fruit? A boo-nana!<\/li>\n\n\n\n
                • Probability of finding it funny:<\/strong> 60%<\/strong><\/li>\n\n\n\n
                • The Breakdown:<\/strong> This one is heavily demographic-dependent. Elementary schoolers will give it a 100% approval rating, while adults will likely just offer a polite, sharp exhale through the nose.<\/li>\n<\/ul>\n\n\n\n

                  3. The Skincare Routine<\/strong><\/p>\n\n\n\n

                    \n
                  • Joke:<\/strong> Why do bananas have to put on sunscreen before they go to the beach? Because they might peel!<\/li>\n\n\n\n
                  • Probability of finding it funny:<\/strong> 70%<\/strong><\/li>\n\n\n\n
                  • The Breakdown:<\/strong> It makes logical sense in the real world and has a crisp punchline. A solid, inoffensive crowd-pleaser for all ages.<\/li>\n<\/ul>\n\n\n\n

                    4. The Visual Gag<\/strong><\/p>\n\n\n\n

                      \n
                    • Joke:<\/strong> What do you call two bananas? A pair of slippers!<\/li>\n\n\n\n
                    • Probability of finding it funny:<\/strong> 75%<\/strong><\/li>\n\n\n\n
                    • The Breakdown:<\/strong> This relies on a great visual gag and a bit of wordplay. It usually gets a genuine chuckle once the imagery clicks in the listener’s head.<\/li>\n<\/ul>\n\n\n\n

                      5. The Traffic Violation<\/strong><\/p>\n\n\n\n

                        \n
                      • Joke:<\/strong> Why did the banana get a ticket? He was caught peeling out of the parking lot!<\/li>\n\n\n\n
                      • Probability of finding it funny:<\/strong> 55%<\/strong><\/li>\n\n\n\n
                      • The Breakdown:<\/strong> A bit more niche. It mostly appeals to people who appreciate mild automotive puns mixed into their daily serving of fruit humor.<\/li>\n<\/ul>\n\n\n\n

                        CLAUDE<\/strong><\/p>\n\n\n\n

                        My favourite AI is Claude, so lets see how Claude fares:<\/p>\n\n\n\n

                        “Tell me 5 jokes about a banana with probabilities”<\/em><\/strong><\/p>\n\n\n\n

                        Claudes answer:<\/p>\n\n\n\n

                        \ud83c\udf4c 5 Banana Jokes \u2014 With Probabilities<\/strong><\/p>\n\n\n\n

                        A carefully researched comedic analysis.<\/p>\n\n\n\n

                        Joke #1
                        Why did the banana go to the doctor? Because it wasn’t peeling well.<\/strong>
                        Probability of a groan: 92%       Probability of a laugh: 61%    <\/p>\n\n\n\n

                        Joke #2
                        What do you call a banana that gets all the girls? A banana smoothie.<\/strong>
                        Probability of confusion: 48%       Probability of a chuckle: 55%   <\/p>\n\n\n\n

                        Joke #3
                        Why did the banana lose the race?
                        It couldn’t ketchup. Wait \u2014 wrong fruit. It slipped up.<\/strong>
                        Probability of eye-roll: 88%       Probability of laughing anyway: 72%  <\/p>\n\n\n\n

                        Joke #4
                        Knock knock. Who’s there? Banana. Banana who? …Banana. (Repeat 4 times, then say:)
                        Orange you glad I didn’t say banana?<\/strong>
                        Probability of a kid losing their mind: 99%       Probability of an adult enjoying it: 23% <\/p>\n\n\n\n

                        Joke #5
                        What’s a banana’s favourite gymnastics move?
                        The splits.<\/strong>
                        Probability of instant regret telling it: 34%       Probability of a genuine laugh: 78%      <\/p>\n\n\n\n

                        THE REASON FOR THE VARIATION IN RESULTS<\/strong><\/p>\n\n\n\n

                        The reason behind the richer variety of banana jokes is a foundational principle of “prompt engineering.” <\/strong>Adding constraints, unusual combinations, and analytical requests fundamentally changes how an AI processes your prompt.<\/p>\n\n\n\n

                        When you ask a Generative AI a layered question instead of a basic one, you aren’t just asking for more<\/em> text; you are changing the underlying mechanics of how the response is generated. Here is exactly why you get richer, better results by altering the prompt:<\/p>\n\n\n\n

                        1. Escaping the “Clich\u00e9 Trap”<\/h3>\n\n\n\n

                        At a fundamental level, Large Language Models (LLMs) like work by predicting the most statistically likely next word.<\/p>\n\n\n\n

                          \n
                        • The Simple Prompt:<\/strong> If you say, “Tell me a joke about a banana,”<\/em> the AI searches for the strongest, most common statistical connection to those words. The result is almost guaranteed to be the most overused dad joke in existence: “Why did the banana go to the doctor? Because it wasn’t peeling well.”<\/em> It is the path of least resistance.
                          <\/li>\n\n\n\n
                        • The Complex Prompt:<\/strong> By asking for 5 jokes and<\/em> a probability of human reception, you force the AI off that well-worn path. It can’t just regurgitate the single most common joke; It has to dig deeper into it’s training data to find five distinct angles, which inherently forces more creativity.<\/li>\n<\/ul>\n\n\n\n

                          2. Expanding the “Attention” Network<\/h3>\n\n\n\n

                          Generative AI uses a mechanism called “Attention” to weigh how different words in your prompt relate to each other. When you add the requirement to calculate probabilities of human reception<\/em>, you activate entirely different conceptual networks within the model. The AI is no longer just looking at [Banana]<\/code> + [Joke]<\/code>. It is now simultaneously cross-referencing [Banana]<\/code> + [Humor Theory]<\/code> + [Statistical Probability]<\/code> + [Human Psychology\/Reception]<\/code>. Weaving these distinct concepts together automatically produces a more nuanced, sophisticated output.<\/p>\n\n\n\n

                          3. Forcing “Chain of Thought” Reasoning<\/h3>\n\n\n\n

                          This is perhaps the most powerful reason the complex prompt works better. By asking the AI to assign a “probability” to how funny the joke is, you are forcing it to evaluate<\/em> the jokes produced.<\/p>\n\n\n\n

                          To give you a probability, it has to implicitly analyze the joke: Is this a pun? Is it absurdism? Does it require specific cultural knowledge?<\/em> Because it now has to justify the joke with a metric, the quality of the joke itself improves. It is effectively checking it’s own work before it is output to ensure the joke matches the analytical rating they are about to give it.<\/p>\n\n\n\n

                          4. Setting the Persona and Formatting<\/h3>\n\n\n\n

                          A simple prompt invites a casual, breezy response. By asking for multiple outputs and a specific analytical data point (probabilities), you are implicitly commanding the AI to adopt a more rigorous, structured persona. It is interpreted as a request for a high-effort, organized response, rather than a quick conversational reply.<\/p>\n\n\n\n

                          IN SUMMARY<\/strong><\/p>\n\n\n\n

                          \n

                          “A simple prompt asks the AI to retrieve a memory. A complex prompt forces the AI to build a customized, multi-dimensional puzzle.”<\/em><\/p>\n<\/blockquote>\n\n\n\n

                          \n

                          What did the banana say after writing a better prompt? “I’m really starting to peel my potential.” Probability of a groan: 91%<\/p>\n<\/blockquote>\n\n\n\n

                          <\/p>\n","protected":false},"excerpt":{"rendered":"

                          “Why did the banana go to the doctor?”Because he wasn’t peeling well! That was the identical answer given when I asked “Tell me a joke about a banana” by ChatGPT, Claude and Gemini. My 4 year old grand-daughter told me that joke, so is AI intelligent? It is not. AI is as dumb as a […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3],"tags":[],"class_list":["post-38","post","type-post","status-publish","format-standard","hentry","category-generative-ai"],"_links":{"self":[{"href":"https:\/\/karloneill.dev\/index.php?rest_route=\/wp\/v2\/posts\/38","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/karloneill.dev\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/karloneill.dev\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/karloneill.dev\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/karloneill.dev\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=38"}],"version-history":[{"count":5,"href":"https:\/\/karloneill.dev\/index.php?rest_route=\/wp\/v2\/posts\/38\/revisions"}],"predecessor-version":[{"id":44,"href":"https:\/\/karloneill.dev\/index.php?rest_route=\/wp\/v2\/posts\/38\/revisions\/44"}],"wp:attachment":[{"href":"https:\/\/karloneill.dev\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=38"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/karloneill.dev\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=38"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/karloneill.dev\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=38"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}